The development of new protective systems must be performed on tools reliable and representative of alive human. In an earlier study, a simplified but realistic modeling of the head-neck-torso system under moderate rear impact was performed. This model of minimum complexity (MC-HNT model) but able to reproduce the 5 first experimental vibration modes was validated in the frequency domain in terms of natural frequencies and damping as well as mode shapes. The human model was then coupled to a car seat-head rest complex on Madymo Code in order to give real body behaviors and accurate T1 accelerations. The hypothesis of linear behavior was used for the torso being subjected to small deformations. The present study shows in detail the methodology carried out for real-world rear impact accident reonstruction in order to establish more accurate neck injury criteria as well as associated tolerance limits. In order to proceed to that, 87 accident cases were simulated using our MC-HNT human body model coupled to 3 Toyota seats under Madymo code. Several injury criteria, such as Neck Fx, Neck Fz, T1 acceleration, NICmax, Nkm and NDC, were calculated in order to correlate the risk of AIS1 neck injury using MCHNT. A similar work has then been done with the BioRID II model. Then a comparison between the predictive risk curves obtained by analyzing the MC-HNT model and the BioRID II model has been performed. This comparison was expressed in terms of Nagelkerke R-square values obtained with these analyses. It appears that the MC-HNT model gives a higher correlation than the BioRID II one for all parameter, and that the lower neck axial force is shown as the best candidate to correlate with the neck injury.